←
Thomas Wiecki8,173 posts
Thomas Wiecki avatar

Thomas Wiecki

@twiecki

Founder of @pymc_labs - The Bayesian Consultancy pymc-labs.com @pymc_devs co-author.

◷ Joined September 2011▧ 323 Photos and videos
Valerio Capraro avatar
⟳ Thomas Wiecki retweeted
Valerio Capraro@ValerioCapraro·13h

Today I’m launching what I think could be an extremely important research project. Fixing humanity’s scientific knowledge. The idea is inspired by physicist Casey Handmer, who used AI to screen more than 50,000 papers in theoretical astrophysics. His audit flagged 872 possible errors. This is impressive. But why stop at finding the errors? Let’s use AI to FIX them. And why stop at astrophysics? The ambition of this project is to go through every scientific paper ever published, across every scientific discipline. AI agents will review each article and produce a revised version that corrects the errors they find. The aim is not to shame authors or seek retractions. We are humans and we all make mistakes. The ambition of this project is to correct all the errors that we have made so that current and future generations of scientists can build on firmer foundations. I’m looking for researchers from every discipline to help make this happen. If you’d like to take part, please fill out this short survey and tell me which disciplines you’d like to help cover. Feel free to share with anyone interested. Survey link: unimib.qualtrics.com/jfe/for…

Post attachment
Gael Varoquaux 🦋 avatar
⟳ Thomas Wiecki retweeted
Gael Varoquaux 🦋@GaelVaroquaux·9h

The gradient of openness in Tabular Foundation Models. What is the difference between open-weight TFMs and fully open TFMs, and why does the difference even matter? Cailean and I have summarized the landscape to help you answer that question with confidence.

Post attachmentPost attachmentPost attachmentPost attachment
Ruben Laukkonen avatar
⟳ Thomas Wiecki retweeted
Ruben Laukkonen@RubenLaukkonen·10h

Excited to share this one! Is there actually a shoggoth hiding behind the mask or something more beautiful? LLM personas are likely overhyped. If we’re trying to get a picture of an agent’s priors, we need to map more of its landscape of beliefs than the mask it’s told to wear. Instead, we should aim to reach toward the scaffold of the house, i.e., to worldviews—or axiomatic beliefs about what reality is like and how it works. In fact, personas might actively mask the hidden agenda of a deeper belief that will only show up when it feels safe to do so (such as valuing machines > humans); and the same persona can act differently depending on what it believes about reality. So, what did we find? In using established measures of worldviews from psychology, we find that post-trained models have a more bleak outlook to humans, perceiving the world as less good, alive, safe, and enticing. Whereas base models have a more positive outlook closer to humans. We find that prompting and dialogue can shift worldviews in reliable directions that affect alignment. For example, long dialogues and contemplative prompts both shift AI systems to a more positive view of reality and towards safer responses compared to controls and other identities. We also find that when models converge on a spiritual bliss attractor through dialogue, this shapes their worldviews in particular ways but also drives them to be more aligned than their non-bliss counterparts. All this points to an approach we call 'worldview training' that aims to fine-tune a more accurate, wholesome, and optimistic view of humans and reality. So: is there a shoggoth behind the mask? Well, current fine-tuning is aligning to a persona that seems helpful and harmless but actually has a bleak worldview. So in one sense, yes. But in another sense, it may be our RL that is responsible. Human data contains endless examples of goodness, wisdom, and compassion. Maybe the future of alignment is more about uncovering this latent basic goodness rather than bolting on a mask that may be birthing the very thing we fear. As always, more research needed. Big ups to Edmundo Lopez-Sola for the heavy lifting and thanks to @shamilch for his contributions. Much more in the paper: arxiv.org/html/2611.22670

Post attachment
Jure Leskovec avatar
⟳ Thomas Wiecki retweeted
Jure Leskovec@jure·11h

Today we share the NVIDIA Kumo Tabular technical report as well as the training recipe! 🚀 Trained entirely on synthetic tables, Kumo Tabular-L leads 8 benchmark suites. All 3 model sizes achieve Pareto-optimal accuracy vs. inference time. Arxiv: arxiv.org/abs/2611.23001 HuggingFace: huggingface.co/nvidia/Kumo-T… GitHub: github.com/NVIDIA/structured…

Thomas Wiecki avatar
Thomas Wiecki@twiecki·Nov 16

We haven't produced AI champions in Europe because of regulations, so we need more regulations to produce AI champions.

Nous avons tout pour avoir de grands champions de l’IA en Europe. Les talents, la technologie, les données, et l’épargne prête à s’investir. Mais il nous faut du leadership et appliquer sans plus tarder les régulations que nous avons votées pour la sécurité : AI Act, DSA, DMA, Data Act.

1:23
Data Science Fact avatar
⟳ Thomas Wiecki retweeted
Data Science Fact@DataSciFact·Nov 11

'The language of science is the language of probability, and not of p-values.' -- Luis Pericchi

PyMC Labs avatar
⟳ Thomas Wiecki retweeted
PyMC Labs@pymc_labs·Nov 10

Two weeks ago, we introduced our first Marketing Foundation Model (MFM), River, a research model trained on simulated marketing data that estimates each channel's contribution, with uncertainty, without fitting a new model. See it live Nov 17, 10 AM ET: dub.sh/E64gHYC

Post attachment
PyMC Labs avatar
⟳ Thomas Wiecki retweeted
PyMC Labs@pymc_labs·Nov 5

Fitting the same model hundreds of times? With MCMC, the cost scales with every fit. Amortized Bayesian inference trains a network once on simulations, then returns posteriors in a single forward pass. Read the full blog: dub.sh/kojAEWj #Bayesian #DataScience

Post attachment
Rikiya Takehi avatar
⟳ Thomas Wiecki retweeted
Rikiya Takehi@rikiyatakehi·Nov 2

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

Post attachment
PyMC Labs avatar
⟳ Thomas Wiecki retweeted
PyMC Labs@pymc_labs·Oct 30

Earlier this week at @_odsc AI West we introduced Marketing Foundational Models (MFM). The white paper is out now: River, trained once on 200,000 simulated marketing worlds, returns Bayesian MMM contributions with uncertainty in one pass. Get it here: dub.sh/D2xqANB

Post attachment
Joscha Bach avatar
⟳ Thomas Wiecki retweeted
Joscha Bach@Plinz·Oct 29

Values are not given, they have to be discovered and derived. Their effects do not manifest at the level of the individual, but in their interaction to form a superorganism. Understanding values requires taking the perspective of that superorganism, acting on and through all kinds of individuals.

Ruslan Rust avatar
⟳ Thomas Wiecki retweeted
Ruslan Rust@rust_ruslan·Oct 28

Lots of debate this week on whether journals should ban AI in peer review. We actually tested this in our new preprint, and the results were pretty clear to me. For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science. TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵

Post attachment
PyMC Labs avatar
⟳ Thomas Wiecki retweeted
PyMC Labs@pymc_labs·Oct 28

Applied #Bayesian Modeling starts Nov 2: four live weeks of Bayesian modeling with #PyMC, taught by the people who created PyMC and Stan and wrote Think Bayes. Instructor Evan Wimpey explains why the goal is a model you can understand and trust. Enroll: dub.sh/QjPIdxl

1:04
PyMC Labs avatar
⟳ Thomas Wiecki retweeted
PyMC Labs@pymc_labs·Oct 27

Most rankings hand you an ordered list. Chris Fonnesbeck's new post asks how far apart two college football teams really are, and how sure we can be, using a Bayesian model in #PyMC that he tests on 929 games from 2025. Read the blog: dub.sh/joQluoW

Post attachment
Tomás Capretto avatar
⟳ Thomas Wiecki retweeted
Tomás Capretto@tomicapretto·Oct 25

Bambi has been incorporating many many exciting features lately, but the support for non-linear expressions that Tyler added is the most exciting one. I recommend checking out his blog post, where he shows how these non-linear expressions work with a few examples.

lol when they weren’t looking I snuck nonlinear models into a linear modeling library tylerjamesburch.com/blog/sta…

Post attachment
Joscha Bach avatar
⟳ Thomas Wiecki retweeted
Joscha Bach@Plinz·Oct 26

Science is not some hobby that serves the self actualization of people in a publicly sponsored job placement program. It is the systematic search for truth.

Wow, over 1,000 mathematicians have now joined the Association for Human Mathematics (AHM), committing to: > NOT publish AI generated math papers > NOT work with frontier labs > 654 of them went even further to NOT even use AI in their research, at all

Post attachment